738 
be obtained. The Taylor method is sensitive to radiometric cali- cognil 
bration errors and quantization effects, especially for water-type 
discriminations. The results, however, for agricultural and forest 
scenes demonstrate that this method will be useful for photointer- of Pu] 
pretation of these regions. omatec 
grams 
run oi 
AUTOMATED CLASSIFICATION SYSTEMS Most ( 
algor: 
Software Systems: There has been a burgeoning development (Craii 
in Canada of software systems for automated classification of MSS systei 
data on CCT's. Within the federal government research on automated requi: 
classification has been carried out by the Department of the Envir- catioi 
onment (Forest Management Institute, Marine Sciences Directorate, that 1 
and Canada Centre for Inland Waters), the Department of National ERTS : 
Defence (Defence Research Board), the Department of Agriculture, and the Lj 
the Department of Energy, Mines, and Resources (Canada Centre for j_ n g a < 
Remote Sensing). Industrially, three Canadian firms are known by 
the authors to have experience in building automated classification 
systems: Computing Devices Company, Ross and Associates, and 0VAAC8. Canad< 
University scientists who are concentrating on automated classifi- compli 
cation as applied to ERTS imagery include: A. Wacker (University of reasoi 
Saskatchewan), H.D. Steiner (University of Waterloo), J. Munday given 
(University of Toronto), G. Rochon (Laval University), E. Langham train: 
(University of Quebec) and W. Davis (University of Alberta). It is data ; 
impractical to attempt to describe each contribution these many selec 
investigators have made to the development of automated analysis of tures 
ERTS imagery. We, instead, will briefly describe two large software j_ n g Q; 
systems, MICA, developed and used at CCRS, and LARSYS, developed at poneni 
Purdue University and used by Wacker at the University of Saskatchewan. theme: 
These two systems are representative of the automated classification can ] D( 
work being carried out in Canada. Tempo: 
Resul- 
The MICA (Modular Interactive Classification Analyzer) class: 
system (Goodenough et a_l 1974) consists of a collection of interact- ved. 
ive programs written in Fortran and assembler languages for operation catioi 
on a PDP-10 with color display, plotter, and electron beam image image 
recorder devices (EBIR). The most important functions of the MICA 
system are listed in Table II. 
are p: 
At CCRS all disk image files and non-image data files are cessii 
in standardized formats to facilitate user and the MICA system book- press« 
keeping. It is thus possible for a user of MICA to examine or modify the h: 
at any time the spectral signature of a class. This is useful if terns : 
one is attempting to determine the nature of the data which prevent to acl 
certain class discriminations. The MICA system is very fast because 
use is made of the fact that a typical ERTS frame contains only 6000 
independent intensity vectors (Shlien and Goodenough 1974). A look 
up scheme is employed to speed classification, yet at the same time 
permits one to easily change the classification algorithm. The MICA 
system is a general purpose multispectral analyzer for pattern re-